Should we use ICA artifact removal in Connectivity analysis?

Hello!
I wanted to ask you about an existential doubt that I have had for a long time:

Should ICA be used to clean artifacts if I am going to perform connectivity analysis?

I'm sure in several posts or maybe Brainstorm tutorials I've read that this introduces a common factor and can inflate the connectivity results.
However I am now unable to find those texts and need some backup. I already found an article that indeed advises not to use ICA or PCA in connectivity analysis.

Thank in advances! :smiley:

PS: I found this:

https://www.researchgate.net/publication/328991449_Does_independent_component_analysis_influence_EEG_connectivity_analyses

What are your thoughts on that?
@Sylvain @Francois @tmedani
:smiley:

Hi @tourette95

ICA is robust in identifying some sources of EEG artifacts, and it is recommended in most cases. However, a portion of the brain signal permanently gets removed when you remove the ICs.

So, it all depends on the quality of your data; I may recommend first visually checking your data and deciding what to do next.

Check also this paper: EEG is better left alone - PubMed

1 Like

I really apreciate your prompt reply :slight_smile:

1 Like